Text-to-image generative artificial intelligence depiction of radiographers and nuclear medicine scientists:a pictorial review

Introduction: Generative artificial intelligence (AI) text-to-image production could reinforce or amplify gender and ethnicity biases1. Text-to-image generative AI tools are used for producing images that represent radiographers and nuclear medicine scientists2. White male stereotyping and masculine cultures can dissuade females and ethnically divergent people from being drawn into a profession3. Evaluation of the diversity of text-to-image AI generated images in radiography and nuclear medicine is warranted. <br/><br/>Method: DALL-E 3, Firefly 2, Stable Diffusion 2.1 and Midjourney 5.2 were utilised to generate a series of individual and group images of radiographers and nuclear medicine scientists. Multiple iterations of images were generated using a variety of prompts. All images were independently analysed by three reviewers for apparent gender, skin tone using the Massey-Martin NIS Skin Colour Scale4, image-to-text alignment and visual reasoning. Discrepancies were resolved by additional reviewer consensus. Ethics was not required. <br/><br/>Results: DALL-E 3, Midjourney 5.2 and Stable Diffusion 2.1 had a higher representation of males and light skin tones than Firefly 2. Image quality metrics were superior for DALL-E 3 and Midjourney 5.2, while inadequate or poor for Firefly 2 and Stable Diffusion 2.1. For both professions there was over-representation of males compared to females. These outcomes are depicted pictorially in this scientific poster. <br/><br/>Conclusion: AI text-to-image generation using DALL-E 3 via GPT-4 has the best overall quality compared to Firefly 2, Midjourney 5.2 and Stable Diffusion 2.1. Nonetheless, DALL-E 3 exhibits inherent biases associated with gender and ethnicity not representative of the diversity of medical radiation practitioners. <br/><br/>Keywords: Artificial intelligence, DALL-E 3, generative AI, GPT-4, radiography, nuclear medicine<br/>References: <br/>1. Currie G, Currie J, Anderson S, &amp; Hewis, J (2024). Gender bias in generative artificial intelligence text-to-image depiction of medical students. Health Education Journal 00:0;1-15.<br/>2. Currie G, Hawk KE and Rohren E (2020). Ethical principles for the application of artificial intelligence (AI) in nuclear medicine and molecular imaging. European Journal of Nuclear Medicine and Molecular Imaging 47: 748–752.<br/>3. Crews DC, Collins CA, Cooper LA. Distinguishing workforce diversity from health equity efforts in medicine. JAMA Health Forum. 2021;2(12):e214820. <br/>4. Massey DS, Martin JA. The NIS Skin Color Scale. Princeton, NJ: Princeton University Press. 2003.<br/>